CAE model local adaptive updating method and system

Abstract

The invention relates to the technical field of computer-aided engineering (CAE) simulation, discloses a local adaptive updating method and system for a CAE model, and aims to solve the problems that solution settings cannot be inherited, computing resources are wasted and grid quality fluctuates due to low full-weight grid efficiency and inconsistent topology in existing parametric simulation. The method comprises the following steps: initializing a geometric model and a local discretization system, and constructing a parameter-node incidence matrix; monitoring parameter changes and positioning a grid influence domain; performing local deformation on the influence domain by using a radial basis function, and executing quality detection and local regeneration; and solving setting is automatically inherited, and topology and physical consistency is verified. Through local adaptive updating, the simulation efficiency and the automation level are remarkably improved, and stable grid quality and lossless inheritance of solution setting are ensured.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided engineering (CAE) simulation technology, specifically relating to a method and system for local adaptive updating of CAE models. Background Technology

[0002] With the continuous evolution of computer-aided engineering technology, numerical simulation analysis of complex industrial products has become a core part of the R&D process in modern manufacturing. In fields such as aerospace, automotive manufacturing, and precision instruments, constructing high-precision numerical models to conduct preliminary research and optimization of physical performance can not only significantly shorten product development cycles but also greatly reduce experimental costs and improve the reliability of the final design.

[0003] Among them, parameterized simulation optimization for geometric features is a key technical path to achieve product iterative evolution. This technology aims to find the optimal design scheme that meets multiple constraints by dynamically adjusting the key structural parameters in the geometric model and combining it with an automated mesh generation and solution process. Its core lies in maintaining a high degree of correlation and synergy among the geometric model, discrete mesh, and physical field description.

[0004] Existing technologies have significant limitations in handling mesh updates during parametric simulations. First, traditional full remeshing methods are inefficient. When dealing with large-scale, complex models, even minor adjustments to geometric parameters often trigger a complete re-execution of geometry cleanup and mesh generation, resulting in extremely high time costs during optimization iterations. Second, the lack of topological consistency between the regenerated mesh and the initial mesh means that previously defined boundary conditions, load application locations, and mesh properties cannot be automatically inherited. Manual rebinding significantly reduces the automation level of the simulation. Third, existing methods suffer from severe computational resource waste in responding to changes in local geometric features. They typically reconstruct the entire model rather than adjusting only the affected areas, lacking accurate identification of parameter influence domains and control over local deformation. Finally, the random perturbations in full remeshing algorithms can easily lead to fluctuations in mesh quality across different iteration steps, causing numerical deviations in simulation results and severely impacting the credibility of parametric studies. These problems collectively constitute the technical bottleneck of model-linked updates during parametric optimization, necessitating a technical solution capable of achieving local adaptive updates of CAE models. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for local adaptive updating of computer-aided engineering models, which can effectively solve the problems mentioned in the background art. Addressing the structural defects in existing technologies during parametric simulation, such as low efficiency of full remeshing, lack of mesh topology consistency leading to the inability to inherit solution settings, severe waste of computational resources caused by local feature changes, and poor mesh quality stability leading to fluctuations in simulation results, this invention establishes a dynamic correlation mechanism between geometric parameters and discrete mesh nodes, achieving accurate identification and local adaptive mesh adjustment only for areas affected by geometric changes.

[0006] To achieve the above objectives, the technical solution adopted by this invention is as follows: a computer-aided engineering model local adaptive update method, comprising the following steps: S1, performing initialization of the geometric model and the locally discretized system, constructing the correlation matrix between geometric parameters and mesh nodes, and storing the initial solution constraint mapping relationship; S2, identifying the changes in geometric parameters through a parameter monitoring mechanism, and determining the set of strongly correlated nodes affected by parameter changes based on the correlation matrix, thereby locating the mesh influence domain; S3, using radial basis functions to perform local deformation updates on the mesh within the influence domain, and performing quality checks and local regeneration on the deformed mesh to ensure that the mesh indicators meet the preset standards; S4, performing automatic inheritance and topology consistency verification of the solution settings, mapping the original physical field constraints and load information to the updated mesh model, and performing simulation pre-verification.

[0007] Preferably, step S1 specifically includes the following steps: S11, digitally encode the key structural parameters in the geometric model through the parameter encoding module, and extract the initial values, accuracy range and motion constraints of the parameters; S12, construct a geometric model and local discretization system management module, calculate the association weight of each grid node with each geometric parameter based on geometric sensitivity analysis or spatial distance weighting algorithm, and generate a large-scale association matrix; S13, establish a constraint mapping library, and associate and store the boundary conditions, load application area and material properties in the initial state with the corresponding grid node identifier or element identifier.

[0008] Preferably, step S2 specifically includes the following steps: S21, capturing dynamic adjustment instructions for geometric parameters in real time and calculating the difference between the parameters in the current iteration step and the previous iteration step; S22, based on the parameter difference and the weight information in the correlation matrix, selecting nodes whose weight values ​​exceed a preset threshold as strongly correlated nodes; S23, performing a topological search outward from the strongly correlated nodes as the core, including the cells containing the strongly correlated nodes into the candidate influence domain, and pruning the candidate influence domain according to spatial curvature or distance constraints, and finally determining the precise range of the mesh influence domain.

[0009] Preferably, step S3 specifically includes the following steps: S31, selecting a predetermined number of control nodes in the geometric feature change region, and calculating the target coordinate position of the control nodes based on the change in geometric parameters; S32, constructing a spatial interpolation model based on radial basis functions, using the displacement of the control nodes as input, and obtaining the interpolation coefficients by solving a system of linear equations; S33, using the interpolation model to calculate the displacement vectors of all non-control nodes in the influence domain, and updating the node coordinates to achieve synchronous mesh deformation; S34, performing geometric distortion checks on the deformed influence domain elements, and performing local remapping or local subdivision for elements that do not meet the preset quality threshold.

[0010] Preferably, step S4 specifically includes the following steps: S41, identifying nodes and elements that have not undergone displacement within the non-influence domain and directly inheriting their corresponding boundary conditions and constraint settings; S42, for areas within the influence domain that have undergone deformation or regeneration, using a spatial topology mapping algorithm to rebind the original load information to the new node or element surface; S43, verifying topological consistency by calculating the total mass, centroid position, and mesh continuity index of key areas of the model before and after the update; S44, running a lightweight pre-simulation process to compare the trend changes in the physical field distribution before and after the update to ensure the physical effectiveness of the model update.

[0011] Preferably, when constructing the association matrix, the geometric model and the local discretization system management module employ a distance-weighted decay function, configured such that the association weight decays non-linearly as the distance from the grid node to the geometric feature center increases. This design ensures that only grids within the physical neighborhood will respond, effectively isolating the impact of local minor adjustments on the global model, thereby fundamentally guaranteeing the absolute stability of grids in non-affected domains.

[0012] Preferably, the parameter encoding module incorporates a multi-level constraint checking mechanism when storing parameter attributes. When the input geometric parameters exceed the preset safe variation range, the system automatically triggers an early warning and truncates the variation to prevent mesh overlap or topology collapse caused by excessive deformation. Furthermore, the encoding module supports the decoupling of complex correlated parameters, transforming multiple interconnected physical dimensions into independent driving variables, thus improving the computational efficiency of the correlation matrix.

[0013] Preferably, the influence domain localization module employs a hierarchical diffusion algorithm when performing topology search. This algorithm starts from strongly correlated nodes and recursively moves layer by layer towards adjacent units until the distance between the searched unit and the geometric feature center exceeds a specific proportion of the feature size, or the unit's correlation weight is lower than a minimum preset value. This dynamic pruning strategy ensures a smooth transition zone between the boundary of the influence domain and the non-influence domain, avoiding significant discontinuities in the mesh gradient during deformation.

[0014] Preferably, the radial basis function employs a multiquadratic surface function with global support characteristics or a compactly supported function with local support characteristics. During interpolation, a specific smoothing factor is introduced to control the smoothness of deformation. The value of the smoothing factor is adaptively adjusted based on the average size of the mesh and a preset step size. This interpolation mechanism ensures that when geometric features are translated, rotated, or scaled, the surrounding mesh can flow naturally like an elastic medium, greatly preserving the distribution characteristics of the original mesh.

[0015] Preferably, the quality inspection step uses various geometric metrics, including but not limited to the Jacobian determinant ratio, element aspect ratio, distortion, and in-plane angle deviation. The system incorporates a hierarchical evaluation standard. For elements with slight distortion, nodal Laplacian smoothing is used for fine-tuning, while for heavily distorted or negative volume elements, a local reconstruction is performed by calling the background mesh generator through a local regeneration module. The boundary of the reconstructed region is strictly constrained within the influence domain to ensure a perfect match with the shared boundary of the external original mesh.

[0016] Preferably, the local subdivision operation employs specific element splitting logic, such as dividing a hexahedral element into eight sub-elements according to a preset ratio, or achieving seamless connection between different refinement levels through a transition element connection strategy. The subdivision process maintains the continuity of the original material properties and maps the physical quantities at the nodes from the coarse mesh to the fine mesh through linear interpolation or higher-order interpolation, ensuring the accuracy compensation of the solution set in the refinement region.

[0017] Preferably, the constraint mapping library supports dynamic rebinding logic. After the mesh update is completed, the system automatically traverses the set of boundary conditions for the affected region. If a boundary condition was originally bound to a specific geometric surface, the system will use a projection algorithm to search for the new mesh surface corresponding to that geometric surface after the update and update the relevant node groups or element groups. This adaptive rebinding mechanism eliminates the need for manual intervention and achieves full automation from geometric modification to solution preparation.

[0018] Preferably, the topology consistency verification is achieved by comparing the global node identifier sequence with the cell connection matrix. For unaffected regions, the identifiers and connections must be 100% consistent; within the affected domain, the historical path of topology evolution is recorded to ensure that newly added nodes or cells are correctly incorporated into the topology of the solution domain. This rigorous consistency verification provides a reliable data foundation for subsequent parametric sensitivity analysis.

[0019] This invention also provides a computer-aided engineering model local adaptive update system for implementing the above-mentioned method, comprising: a parameter encoding and initialization module for digitally defining geometric parameters and constructing an initial correlation matrix and constraint mapping library; a parameter monitoring and influence domain location module for capturing parameter changes and determining the affected local mesh regions; a local deformation and quality optimization module configured to drive mesh deformation using radial basis functions and perform local quality detection and remapping; and a solution setting inheritance and verification module for automatically mapping boundary conditions and loads and verifying the topological and physical validity of the updated model.

[0020] Preferably, the parameter monitoring and influence domain localization module includes a high-frequency sampler, which can capture parameter commands sent by the external geometry engine at extremely short time intervals and trigger the recalculation of associated weights in real time. This module uses an asynchronous parallel processing architecture to complete the influence domain filtering in a background thread, thereby ensuring the smoothness of the user interface.

[0021] Preferably, the local deformation and quality optimization module integrates a multi-core parallel acceleration algorithm. When calculating the interpolation coefficients and nodal displacements of a large-scale influence domain, it can fully utilize the computing power of the multi-core processor, keeping the deformation time within the range of seconds. Simultaneously, this module is also equipped with a preset operator library, which can automatically select the optimal deformation strategy based on different element types (such as tetrahedral, hexahedral, or shell elements).

[0022] Preferably, the solver settings inheritance and verification module interfaces with mainstream computer-aided engineering solvers through a standardized data interface. It can read and modify the solver's input file structure, directly writing the updated coordinate information and topological relationships into the corresponding input cards, thus achieving closed-loop integration of model updates and numerical simulations.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention overcomes the efficiency bottleneck of full remeshing in parametric simulation by constructing a linkage mechanism between the geometric model and the locally discretized system. First, the establishment of the correlation matrix achieves a precise logical mapping between geometric driving variables and underlying mesh nodes, enabling the system to identify affected local regions with minimal computational overhead, avoiding redundant geometric cleanup and mesh generation on the overall model. This localized processing mode reduces the model update process, which originally took hours, to seconds or minutes, significantly improving the efficiency of simulation tasks requiring frequent iterations, such as multivariate optimization and tolerance analysis.

[0024] Furthermore, the radial basis function deformation technique and local quality optimization strategy introduced in this invention, while ensuring local mesh updates, maximize the maintenance of the initial mesh's topological connectivity and distribution characteristics. This topological consistency not only guarantees the stability of mesh quality and reduces simulation result fluctuations caused by mesh perturbations, but more importantly, it achieves lossless inheritance of solution parameters such as boundary conditions and load settings. Through automated mapping and rebinding, the cumbersome manual operation process in parametric iteration is eliminated, significantly enhancing the automation level of the simulation process and the reliability of the results, laying a solid technical foundation for achieving true design-simulation integration.

[0025] Furthermore, by using a distance-weighted correlation matrix construction method, the boundary of the influence domain can be accurately delineated in an adaptive manner, achieving optimal allocation of computing resources. Since only 5% to 10% of the affected area is processed, while keeping the remaining 90% or more of the model frozen, this strategy fundamentally solves the problem of wasted global computation caused by fine-tuning local features in large-scale complex industrial models.

[0026] Furthermore, by performing spatial interpolation and node coordinate updates based on radial basis functions, a smooth transition in the mesh deformation process is achieved. Combined with a local subdivision and regeneration mechanism for distorted elements, this invention can consistently control the mesh quality within a preset safety threshold under various harsh parameter variations, ensuring the convergence and numerical accuracy of subsequent finite element analysis or finite volume method solutions.

[0027] Furthermore, by integrating solver setting inheritance and topology verification functions, this invention establishes a logical closed loop from geometric input to solver preparation. In the topology consistency verification stage, by comparing the identifier sequence with the physical field distribution trend, logical errors or physical discontinuities that may occur during model updates are effectively eliminated, thus improving the robustness of the system.

[0028] Furthermore, the modular system architecture and standard data interface design adopted in this invention give it strong scalability and compatibility. Whether for detailed analysis of spacecraft structural components or collision optimization of automobiles, this system can provide stable technical support, significantly reducing the computing power and time costs for enterprises in the development of complex products, and has significant industrial application value. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall technical architecture of a local adaptive update method for CAE models proposed in this invention. Detailed Implementation

[0030] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0031] Example 1 In the automotive industry's R&D system, lightweighting and stiffness optimization of body panels such as the engine hood are core aspects. This embodiment uses the parametric optimization of engine hood bolt holes as an example to explain in detail the specific implementation logic of a computer-aided engineering model local adaptive update method provided by this invention.

[0032] In this embodiment, the initial computer-aided engineering model is a car engine hood, with physical dimensions set to 1500 mm in length, 1200 mm in width, and 1.5 mm in thickness. The model has bolt holes at four key support locations, with the initial x-coordinate parameter of the first bolt hole defined as 100 mm and the initial radius parameter as 5 mm. The discretization result of the model is a high-quality hexahedral mesh containing a total of 120,000 nodes and 100,000 elements. In the initial simulation environment configuration, the circumferential edges of the engine hood are set to fully fixed constraints, and a concentrated force load of 10 kN is applied to the inner circumference of the bolt holes. The hardware execution environment of this embodiment is based on a multi-core processor architecture, equipped with 32 GB of synchronous dynamic random access memory and a high-speed solid-state drive, while the software environment runs on a professional simulation integration platform under a 64-bit operating system.

[0033] Step S1 is executed to initialize the geometric model and the locally discretized system. In this step, the parameter encoding module is first activated to digitally encode the key structural parameters in the geometric model. The parameter encoding module not only records the numerical information of Pos_Hole1_X as 100 mm and Dim_Hole1_R as 5 mm, but also simultaneously extracts the floating-point precision requirements for these parameters, i.e., the precision must reach ±0.01 mm. At the same time, the encoding module defines the motion constraint range for the parameters, stipulating that the variation range of the horizontal coordinate is strictly limited to between 95 mm and 105 mm to ensure the engineering feasibility of the structure.

[0034] Where Pos_Hole1_X is the position parameter, and the encoding rule is: Pos_<feature name>_<axis>; Dim_Hole1_R is the size parameter, and the encoding rule is: Dim_<feature name>_<type>.

[0035] The adjustable parameters of the CAE model are standardized and encoded into "parameter-attribute-constraint" triples, which are then stored as structured data, as shown in Table 1. Table 1 Next, a geometric model and local discretization system management module is constructed. Its core task is to calculate the association weights between geometric parameters and mesh nodes. The management module traverses all 120,000 nodes, applying a spatial distance weighting algorithm to calculate the physical distance of each node relative to the geometric feature center of the bolt hole. For the horizontal coordinate parameter Pos_Hole1_X, the management module identifies all nodes within a 10 mm radius of the hole center and assigns association weights based on the reciprocal of the distance. When the weight value is greater than or equal to 0.1, these nodes are marked as strongly associated nodes; in this embodiment, nodes numbered between 5001 and 5120 are included in this category. Finally, the system generates a large-scale association matrix with 120,000 rows and 2 columns, and stores it in a high-speed cache in memory.

[0036] Subsequently, a constraint mapping library was established. This library serves to solidify the physical boundary conditions with the underlying discrete identifiers. Specifically, the system binds the fixed constraints on the engine hood edge to a sequence of nodes numbered 1 to 500, while associating the 10 kN concentrated force load with nodes 5051 to 5080 on the bolt hole edge. This mapping relationship ensures that during subsequent mesh deformation, the load and constraints can be accurately located to their corresponding physical positions.

[0037] Step S2 is executed, where the change in geometric parameters is identified through a parameter monitoring mechanism. In the iterative optimization process of this embodiment, the designer issues an instruction to adjust the horizontal coordinate of the bolt hole from 100 mm to 105 mm and the hole radius from 5 mm to 6 mm. The parameter monitoring module frequently captures this dynamic adjustment instruction and, through subtraction, calculates that the difference in the horizontal coordinate is 5 mm and the difference in the radius is 1 mm.

[0038] Based on the pre-constructed correlation matrix, the influence domain localization module quickly activates. This module filters out the set of nodes most affected by the parameter change, namely nodes 5001 to 5120, according to the rule that the correlation weight value exceeds a preset threshold. To ensure the continuity of deformation, the localization module uses these strongly correlated nodes as the core and expands the search outwards along the topology. The system automatically identifies hexahedral elements containing these nodes and includes them in the candidate influence domain, initially numbered from 4001 to 4100. To optimize computational resources, the system further introduces spatial curvature constraints to fine-trim the candidate influence domain. By determining the Euclidean distance from the node to the hole center, nodes with a distance exceeding 12 mm are excluded, ultimately locking in a precise mesh influence domain containing nodes 5001 to 5100 and elements 4001 to 4080.

[0039] Step S3 is executed, where the mesh within the influence domain is locally deformed and updated using radial basis functions. First, control nodes are selected at the characteristic edges of the bolt holes. In this embodiment, the center node 5050 and edge nodes 5051 to 5080 are selected. Based on the parameter changes, the new coordinate positions of these control nodes are calculated; for example, the target abscissa of the center node is set to 105 mm.

[0040] Subsequently, a spatial interpolation model based on radial basis functions was constructed. This model uses the known displacement vectors of the control nodes as input and selects a multi-quadratic surface function with global support properties as the basis functions. By solving a system of linear equations, the system obtains a set of interpolation coefficients reflecting the spatial deformation trend. Using these coefficients, the interpolation model can calculate the displacement vectors of all non-control nodes within the influence domain. By updating the coordinates of these nodes, the mesh achieves smooth deformation synchronized with the geometric features; the bolt holes on the hood smoothly move to their new positions and complete the hole enlargement process.

[0041] After deformation, the system immediately enters the quality inspection and local regeneration phase. The inspection module uses geometric metrics such as the Jacobian determinant ratio and element aspect ratio to scan each deformed element. In this embodiment, elements numbered 4050 to 4055 were found to have a large displacement, resulting in an aspect ratio of 3.8, exceeding the preset value of 3. To address this slight distortion, the system triggers local subdivision logic, splitting these hexahedral elements into smaller sub-elements according to an octree structure. The aspect ratio of the subdivided new elements is reduced to 2.5, fully meeting the preset quality standard, ensuring that the overall mesh pass rate remains above 99.2%.

[0042] Step S4 is executed to perform automatic inheritance and topology consistency verification of the solution settings. First, stable regions within the non-influence domain are identified. Since edge constraint nodes numbered 1 to 500 are in the non-influence domain, their displacement increments are zero, and the system directly inherits their initial fixed constraint settings. For regions within the influence domain that undergo deformation and subdivision, the system uses a spatial topology mapping algorithm. The original concentrated loads in the bolt holes are positioned by projection and automatically re-bound from the old edge node sequence to the new subdivided node sequence, i.e., nodes 7001 to 7024, while the load values ​​and directions remain unchanged.

[0043] To ensure the logical rigor of the model update, topological consistency verification was performed. The system compared the global node identifier sequences before and after the update to verify whether the boundary connections between the affected and unaffected domains still satisfied the shared edge constraints. The results showed that all shared edges achieved a 100% matching rate, and there were no dangling nodes or overlapping elements. Finally, a lightweight pre-simulation process was run. In this process, the system performed a simplified linear static analysis on the updated model, calculating a maximum stress value of 122 MPa. Compared with the initial model's 120 MPa, the stress distribution trend was highly consistent, with a deviation ratio of only 1.7%, far below the 3% warning threshold, thus verifying the physical effectiveness of the model update.

[0044] Example 2 In the field of aero-engine development, the distribution and shape of cooling holes on turbine blades have a decisive influence on the high-temperature resistance of the blades. This embodiment takes the optimization of the layout of multiple rows of cooling holes on an aero-turbine blade as an example to illustrate the application of the present invention in scenarios with complex curved surfaces and high precision requirements.

[0045] Aero-engine turbine blades have highly complex twisted shapes with hundreds of tiny cooling holes distributed on their surfaces. The initial model contains approximately 2 million tetrahedral mesh elements. The goal of parametric optimization is to adjust the tilt angle and aperture size of the second row of cooling holes to optimize the coverage efficiency of the cooling airflow.

[0046] In step S1, the initialization module first digitally encodes parameters such as the geometric center, axis vector, and outlet diameter of each cooling hole. Because the cooling holes are located in a high-temperature, high-pressure flow channel environment, the parameter encoding module introduces a more stringent multi-level constraint check mechanism to prevent structural strength failure due to excessively large hole diameters. Subsequently, the geometric model and local discretization system management module calculates the association weights of millions of mesh nodes relative to the cooling hole axis based on geometric sensitivity analysis. Due to the extremely high discrete density of tetrahedral meshes in curved regions, the system employs a nonlinear decay function based on distance weights to ensure that the influence range of the weights is strictly limited to a region three times the hole diameter around the hole.

[0047] In step S2, when the design parameters require an increase of 5 degrees in the tilt angle of the cooling holes, the parameter monitoring module captures this angle change in real time. The influence domain location module identifies the affected local node set through the correlation matrix. Due to the large curvature of the blade surface, the location module employs a hierarchical diffusion algorithm when performing topology search. This algorithm starts from the hole wall nodes and recursively moves layer by layer towards the outer elements until the correlation weight of the elements is lower than a preset minimum value, thereby outlining a precise funnel-shaped influence domain on the complex tortuous surface.

[0048] Execute step S3, and drive local deformation using radial basis functions. In the thin-walled structure of the turbine blade, deformation can easily cause mesh penetration. Therefore, the interpolation model introduces a specific smoothing factor, which is adaptively adjusted according to the average edge length of the tetrahedral elements. By solving the interpolation coefficients, the system drives the hole wall mesh to rotate at a new angle. In the quality inspection link, the system focuses on monitoring the distortion of the tetrahedral elements. If it is found that some elements near the inner cavity of the blade show a serious negative volume trend, the local regeneration module will call the background mesh generator to perform local re-tetrahedralization on the affected local cavity and ensure that the boundary of the reconstructed area is topologically stitched with the main body mesh of the blade.

[0049] Execute step S4. The solution setting inheritance module faces complex boundary conditions of fluid-structure interaction. The system uses dynamic re-binding logic to automatically map the inlet pressure load of the cooling air flow and the outlet heat convection coefficient to the deformed hole wall surface. The topological consistency verification ensures that the newly added reconstructed nodes are correctly incorporated into the solution domain by recording the historical path of topological evolution. The pre-simulation link ensures the continuity of high-fidelity simulation by comparing the velocity vector distribution trends at the outlet of the cooling hole before and after the update, ensuring that the adaptive update of the local mesh does not introduce spurious physical fluctuations.

[0050] Embodiment 3 In the precision electronic device industry, the internal structural components of the mobile phone middle frame usually contain a large number of micro features, such as screw posts, buckles, and antenna brackets. This embodiment takes the sensitivity analysis scenario of the signal occlusion effect caused by the displacement of the antenna bracket in the mobile phone middle frame as an example to illustrate the implementation process of the present invention in a multi-component and multi-parameter linkage environment.

[0051] The mobile phone middle frame model is a large-scale assembly integrating various material properties. The installation position Pos_Antenna of the antenna bracket needs to be finely adjusted frequently according to the radio frequency simulation results. The initial model uses a hybrid mesh technology, with high-precision hexahedral meshes in the antenna area and transitional pentahedral or tetrahedral meshes in the remaining structures.

[0052] Execute step S1. The parameter encoding module supports the decoupling of complex associated parameters, converting the translation and rotation of the antenna bracket in three-dimensional space into independent driving variables. When constructing the association matrix, the management module not only considers the spatial distance but also introduces the material interface damping coefficient to prevent non-physical perturbations of non-connected components caused by cross-part mesh deformation. The constraint mapping library stores the contact pair information between the antenna bracket and the middle frame, electromagnetic shielding constraints, and the node identifiers of the grounding supports.

[0053] In step S2, the parameter monitoring module captures micron-level adjustment commands for the antenna support position at extremely short time intervals. Due to the extremely small gap between the support and surrounding structural components, the influence domain localization module, through an asynchronous parallel processing architecture, rapidly completes the filtering and trimming of the influence domain in a background thread. The trimming strategy ensures that the boundary of the influence domain always falls within a single material, avoiding mesh distortion across the contact interface.

[0054] In step S3, the local deformation module integrates a multi-core parallel acceleration algorithm. When processing the high-density mesh in the antenna region, the system distributes the node displacement calculation task to multiple physical cores of the processor, significantly reducing interpolation time. For potential interference phenomena after deformation, the quality inspection step introduces a collision interference check. If spatial overlap between the support mesh and the mid-frame mesh is detected, the system automatically triggers a Laplace smoothing algorithm to fine-tune the nodes and eliminate geometric distortion.

[0055] In step S4, the solver setup inheritance module interfaces with the electromagnetic simulation solver using a standardized data interface. It reads the new coordinates after the antenna position change and automatically updates the node coordinate segments in the solver input card. Topology consistency verification calculates the total mass and centroid position of the model before and after the update to ensure there is no non-physical loss of material density due to mesh reconstruction. Lightweight pre-simulation compares the electric field distribution trend in the near field of the antenna to confirm that the model update fully reflects the changes in the physical design, providing a reliable basic model for subsequent full-wave electromagnetic analysis.

[0056] In the above embodiments, the operational logic of the geometric model and the local discretization system management module is the core support for achieving local adaptive updates. This module is configured to follow a physical distribution logic based on the principle of energy conservation when performing weight calculations. That is, when a geometric parameter is fine-tuned, the resulting geometric strain energy should decrease exponentially with increasing distance from the feature center. This logic can be described in simple terms as follows: obtain the difference vector between the target node's coordinates and the feature center's coordinates; calculate the length of this difference vector to obtain the physical distance; pass this physical distance as input to a preset attenuation operator; the value output by this operator is the associated weight of the node under that parameter.

[0057] Furthermore, the parameter encoding module possesses an automatic decoupling function when handling multi-variable linkages. For example, when the length and width of a rectangular slot change simultaneously, the encoding module decomposes this complex change into the translational motion of four boundary lines and assigns an independent parameter sequence to each boundary translation. This design allows the correlation matrix to be stored in the form of a sparse matrix, greatly saving memory addressing space overhead and improving access speed in large-scale industrial models.

[0058] In the logical framework of the influence domain localization module, the hierarchical diffusion algorithm is crucial for ensuring deformation stability. During execution, the algorithm first marks all elements directly in contact with geometric features as first-level influence elements. Then, it searches for all neighboring elements sharing nodes with first-level elements and marks them as second-level. The system presets either a total level threshold or a weight accumulation decay threshold. When the diffusion search reaches the preset boundary conditions, the system stops searching and performs a transition protection process on the outermost elements. This transition protection, by assigning smaller deformation weights to edge nodes, maintains first- or second-order continuity in the displacement gradient between the influence domain and non-influence domains, effectively eliminating stress concentration phenomena that may occur after mesh deformation.

[0059] In the implementation details of the local deformation and quality optimization module, the selection of radial basis functions is highly flexible. For large displacement deformation scenarios, the system prioritizes functions with global support properties to ensure the smoothness of the deformation field across the entire influence domain. For local regions containing numerous micro-holes or fine features, the system switches to tightly supported functions with local support properties to prevent deformation of one feature from causing topological collapse of neighboring features. During the solution of the linear equations, to cope with the computational pressure brought by a large number of nodes, the system employs the preconditional conjugate gradient method. This method obtains interpolation coefficients through iterative approximation, avoiding the huge memory consumption of directly inverting the matrix, enabling second-level model update response even on ordinary mobile workstations.

[0060] For the quality inspection steps, this invention establishes a multi-dimensional quantitative evaluation system. The Jacobian determinant ratio is used to evaluate whether the volume mapping of the element is faithful, the aspect ratio is used to evaluate excessive stretching of the element in a specific direction, and the distortion is used to evaluate the flatness of the element surface. The system's built-in hierarchical evaluation criteria stipulate that: if the index is within the safe range, no operation is performed; if it is within the warning range, optimization is performed using a smoothing technique that moves local node positions; if it is within the failure range (e.g., negative volume occurs), the region must be removed through a local regeneration module and the mesh refilled. This hierarchical processing logic ensures mesh quality while preserving the topological information of the original mesh to the maximum extent, achieving an optimal balance between computational overhead and model accuracy.

[0061] In the inheritance logic of the solution setup, the spatial topology mapping algorithm acts as a data bridge. During execution, the algorithm first establishes a search index based on spatial coordinates. When a new node is generated, the algorithm searches for its projection position in the initial model and performs weighted interpolation based on the surrounding physical field information (such as load distribution and constraint states) to assign physical properties to the new node. This adaptive rebinding mechanism is particularly suitable for parametric optimization tasks requiring hundreds or thousands of iterations because it completely replaces tedious, repetitive manual labor.

[0062] Furthermore, the topology consistency verification step is not only a checking procedure but also a data error correction mechanism. During the verification process, if the system detects mesh mismatches (such as gaps or overlaps) at the influence domain boundary, the verification module automatically sends a feedback signal to the deformation module, requesting a readjustment of the smoothing factor or a redefined influence domain boundary until 100% consistency is met. This closed-loop control design fundamentally ensures the robustness of the CAE model's local adaptive update system.

[0063] like Figure 1 This invention proposes a computer-aided engineering model local adaptive update system, in which various modules exchange data via a high-speed internal bus. The parameter encoding and initialization module acts as the system's data source, transforming complex geometric information into a computer-recognizable logical sequence. The parameter monitoring and influence domain location module serves as the system's trigger, sensing external design intent in real time through a high-frequency sampler. The local deformation and quality optimization module is the heart of the system, utilizing efficient numerical algorithms to map from geometric space to physical discrete space. The solution setting inheritance and verification module acts as the system's quality gate, ensuring that each updated model can be directly used for solution calculation.

[0064] This systematic architecture design enables the invention not only to handle simple part optimization but also to handle real-time adaptive updates of large and complex assemblies. When processing vehicle collision models or full-engine structural models containing tens of millions of degrees of freedom, the system can improve model update efficiency by more than two orders of magnitude by processing only less than 5% of the locally affected areas. Because it preserves most of the original mesh topology, the consistency of simulation results is significantly improved, which is invaluable for tolerance design and robustness optimization requiring analysis of minute parameter fluctuations.

[0065] Furthermore, this system boasts excellent cross-platform compatibility, enabling seamless data transfer with mainstream CAD and CAE software via standardized data interfaces such as JSON or XML. In practical industrial applications, this system can be integrated into existing simulation workflow management platforms, serving as the underlying automation engine to support a fully automated closed-loop process from geometric changes to simulation evaluation.

[0066] Through the above three examples from different dimensions, it can be seen that the present invention demonstrates excellent adaptability and extremely high computational efficiency when dealing with CAE model update tasks of different physical domains, different mesh types, and different complexity levels. By establishing a dynamic correlation mechanism between geometric parameters and discrete nodes, the present invention completely solves the efficiency and consistency problems caused by full remeshing, providing a solid technical guarantee for high-frequency, high-efficiency, and automated simulation-driven design of modern industrial products.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for local adaptive updating of a CAE model, characterized in that, Includes the following steps: Step S1: Initialize the geometric model and the locally discretized system, construct the correlation matrix between geometric parameters and mesh nodes, and store the initial solution constraint mapping relationship; Step S2: Identify the changes in geometric parameters through the parameter monitoring mechanism, and determine the set of strongly correlated nodes affected by parameter changes based on the correlation matrix, thereby locating the grid influence domain; Step S3: Use radial basis functions to perform local deformation updates on the mesh within the influence domain; Step S2 specifically includes the following steps: S21. Real-time capture of dynamic adjustment instructions for geometric parameters, and calculation of the difference between the parameters in the current iteration step and the previous iteration step; S22. Based on the parameter difference and the weight information in the correlation matrix, select nodes whose weight values ​​exceed a preset threshold as strongly correlated nodes. S23. Perform a topology search outward from the core of strongly correlated nodes, include the cells containing strongly correlated nodes into the candidate influence domain, and prune the candidate influence domain according to spatial curvature or distance constraints to finally determine the precise range of the grid influence domain.

2. The method according to claim 1, characterized in that, Step S1 specifically includes the following steps: S11. The key structural parameters in the geometric model are digitally encoded through the parameter encoding module, and the initial values, accuracy range and motion constraints of the parameters are extracted. S12. Construct a geometric model and a locally discretized system management module. Calculate the correlation weight of each grid node with respect to each geometric parameter based on geometric sensitivity analysis or spatial distance weighting algorithm, and generate a large-scale correlation matrix. S13. Establish a constraint mapping library to associate and store the boundary conditions, load application areas, and material properties under the initial state with the corresponding mesh node identifiers or element identifiers.

3. The method according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31, selecting a predetermined number of control nodes in the geometric feature change region, and calculating the target coordinate position of the control nodes based on the change in geometric parameters; S32, constructing a spatial interpolation model based on radial basis functions, using the displacement of the control nodes as input, and obtaining the interpolation coefficients by solving a system of linear equations; S33, using the interpolation model to calculate the displacement vector of all non-control nodes in the influence domain, and performing node coordinate updates to achieve synchronous mesh deformation.

4. The method according to claim 3, characterized in that, Step S3 further includes: S34, performing geometric distortion checks on the deformed influence domain units, and performing local re-division or local subdivision on units that do not meet the preset quality threshold.

5. The method according to claim 1, characterized in that, The method further includes: Step S4: Perform automatic inheritance and topology consistency verification of the solution settings, map the original physical field constraints and load information to the updated mesh model, and perform simulation pre-verification.

6. The method according to claim 5, characterized in that, Step S4 specifically includes the following steps: S41. Identify nodes and elements that have not undergone displacement within the non-influence domain and directly inherit their corresponding boundary conditions and constraint settings; S42. For areas within the influence domain that have deformed or regenerated, the original load information is re-bound to new nodes or element surfaces using a spatial topology mapping algorithm. S43. Verify topological consistency by calculating the total mass, centroid position, and mesh continuity index of the key regions of the model before and after the update; S44. Run the lightweight pre-simulation process to compare the trend changes of the physical field distribution before and after the update, and ensure the physical effectiveness of the model update.

7. A computer-aided engineering model local adaptive update system, used to implement the method described in claims 1-6, characterized in that, The system includes: The parameter encoding and initialization module is used to digitally define geometric parameters and build an initial correlation matrix and constraint mapping library; The parameter monitoring and influence domain location module is used to capture parameter changes and determine the affected local grid areas; The local deformation and quality optimization module is configured to drive mesh deformation using radial basis functions and perform local quality checks and rewiring.

8. The system according to claim 7, characterized in that, The system also includes a solver setting inheritance and verification module, which is used to complete the automatic mapping of boundary conditions and loads, as well as the verification of the topological and physical validity of the updated model.

Images